This study analyses 1400 research projects of the top 20 R&D-spending pharmaceuticals to identify the determinants of successful research projects. We provide clear evidence that externally sourced projects and projects involving biotechnologies perform better than internal projects and chemical projects, respectively. Controlling for these effects, we find that big pharma should either build a critical mass of disease area knowledge or diversify projects over different DA’s in order to obtain higher success probabilities. Projects in which a firm has built a critical mass of disease knowledge (through at least 10 projects per DA) are significantly more likely to reach clinical testing. Moreover, within large disease areas, the success probabilities of internal projects increases when a few (less than 20%) externally sourced projects are involved. We interpret this finding as knowledge spillovers from external to internal projects, as the limited number of external projects enables the same people to be involved in both external and internal research projects and apply externally generated knowledge internally.